Artificial intelligence is in fashion these days, as is its energy consumption. From being an innovative technology that would usher in a new era, AI has recently become a problem for many: a problem that makes electricity more expensive and, it seems, does not really live up to the promise of its developers. Big tech companies continue to promise tens of billions in AI investments. Now, it may have to spend a larger portion of those billions on securing its own power and proving itself worth the money AI users would have to spend on the technology.
Recently, a growing number of commentators have questioned the point of incorporating AI into corporate operations. Bloomberg analysts argued this week that the entire AI story shows signs of overvaluation; Uber’s chief operating officer admitted that the company’s investment in technology has not led to the expected productivity gains. Meanwhile, cities are banning data centers across the United States. Because they are driving up electricity prices, using a lot of water and threatening the quality of life in the areas where they are built, according to a recent Gallup poll.
The energy issue seems to be especially delicate. The energy consumption of data centers hosting artificial intelligence technology is already noticeable. Analysts compare this consumption with that of entire nations. However, some have argued that higher electricity prices in some locations with many data centers are not directly related to their electricity consumption but rather to the local energy mix. If that mix includes a lot of wind and/or solar power, bills tend to be higher (as is the case in Europe) and the presence of data centers simply exacerbates an already existing problem.
Related: Strait of Hormuz may reopen, but the system is already broken
This is not something that local communities seem to worry much about, and rightly so. The International Energy Agency said in a report in April this year that electricity demand from data centers overall last year increased 17% from the previous year, and demand from data centers hosting AI increased even more substantially. The report noted that while energy consumption per AI task is declining, the increasing use of AI is offsetting this decline, and then some. In short, AI is a waste of energy.
Wood Mackenzie reported this week that political opposition to AI developers is rising across party lines, as politicians address local community concerns with their respective legislative authorities, focusing on making Big Tech pay out of pocket for the investments needed to secure the electricity and transmission infrastructure data centers need.
“The growth in electricity demand, including demand for new data centers, has so far not had much impact on energy prices in the United States,” Wood Mac noted in its report. “However, it is starting to become a major factor in some areas, including the PJM network that stretches from New Jersey to Tennessee.” It seems irrelevant whether fears of higher electricity prices have a basis in actual prices: those fears are driving actions against data centers. Network restrictions don’t help.
The Wall Street Journal wrote this week that Big Tech is delaying its plans for new data centers, due to “supply chain delays, permitting fights and availability of power supplies,” among other factors slowing the data center boom. The availability of electricity supply is directly related to the price of that electricity. Securing the electricity supply needed for consumers as large as data centers is also related to the price of electricity: PJM Interconnection said last month that it would need to invest an additional $23.1 billion in the area in which it operates, which is the largest in the United States. Those billions will have to be “shared,” so to speak, and the question is who they will be shared with: just the data center operators or everyone in the PJM area.
Meanwhile, Big Tech is addressing its growing electricity consumption problem in a way suggested by some in political and grid operating circles: generating its own electricity. In fact, Big Tech is adding billions of dollars to its spending plans to build its own generation facilities. Of course, this is raising questions about the profitability of the whole AI rush, but it seems to be the most realistic approach to dealing with at least some of the local community and political opposition to the technology.
As for the question of what these generating facilities will run on, the answer, to the chagrin of many, would be oil and gas; in the United States, mainly gas, but also nuclear energy. Without them, fears of an electricity shortage could become reality, and the prices that accompany it could become a reality. It is doubtful that anyone in a position with decision-making power would allow this to happen. And this means that big technology companies would have to “bring their own electricity.”
By Irina Slav for Oilprice.com
More Oilprice.com Top Reads
Oilprice Intelligence gives you the signals before they become headline news. This is the same expert analysis read by veteran traders and political advisors. Get it for free, twice a week, and you’ll always know why the market moves before everyone else.
You’ll get the geopolitical intelligence, hidden inventory data, and market rumors that move billions, and we’ll send you $389 in premium energy intelligence, on us, just for subscribing. Join over 400,000 readers today. Get access immediately by clicking here.
